==PROF== Kernoscope 0.1.0 - provenance-honest profile (no performance counters required)
==PROF== Target: bench/artifacts/.build/ledger_probe --kernel norm --mb 256 --target-s 1.2
==PROF== DRAM 估算 909783.5 MB（显存忙碌 71.4% × 峰值 909 GB/s，窗口覆盖率 71%） → 473.11 MB/launch
==PROF== Kernel time: 1228.744 ms of 4192.466 ms wall (29.3%)
==PROF== Window: mode=auto（已裁剪） 采样 7 点 / 参与积分 1.4 s / 裁掉 2.8 s / 进程 wall 4.2 s
==PROF== source[kernel activity]: nsys sqlite CUPTI_ACTIVITY_KIND_KERNEL (no counters needed)
==PROF== source[device counters]: pynvml (energy/PCIe/mem-busy are management counters, available without admin)
==PROF== source[static]: not used
==PROF== source[peaks]: measured calibration (bench/artifacts/20261008-225555/machine-profile.json)

Kernel: norm_k(const float *, const float *, float *, long)
  nsys short name: norm_k
  1923 launch(es), 1228.7441 ms total, 0.6390 ms mean

  Section: Duration                           [exact]
    gpu__time_duration.sum                                1228.74 ms       [exact]
    ks__launch_time_mean                                    0.639 ms       [exact]
    ks__launch_time_min                                    0.6327 ms       [exact]
    ks__launch_time_max                                    0.6601 ms       [exact]
    ks__launch_count                                         1923          [exact]

  Section: LaunchStats                        [exact]
    launch__grid_size                                        1024          [exact]
    launch__block_size                                        256          [exact]
    launch__thread_count                                   262144          [exact]
    launch__registers_per_thread                               34          [exact]
    ks__registers_used_for_occupancy                           34          [exact]
    launch__shared_mem_per_block_static                         0 B        [exact]
    launch__shared_mem_per_block_dynamic                        0 B        [exact]
    launch__local_mem_per_thread                                0 B        [exact]
    ks__cuda_graph_capture                                    yes          [exact]

  Section: Occupancy                          [exact] (+1 n/a)
    ks__theoretical_occupancy_pct                             100 %        [exact]
    ks__blocks_per_sm                                           6          [exact]
    ks__warps_per_sm                                           48          [exact]
    ks__occupancy_limiter                          registers+threads+warps          [exact]  (并列的限制因素用 + 连接)
    sm__maximum_warps_per_active_cycle_pct                    n/a          [unavailable]  requires GPU hardware performance counters, which are restricted here (RmProfilingAdminOnly=1 and no CAP_SYS_ADMIN/CAP_PERFMON in the container); CUPTI reports 

  Section: SpeedOfLight                       [unavailable] (+2 n/a)
    est__achieved_tflops                                      n/a          [unavailable]  workload not provided: pass --ks-gemm M,N,K or --ks-flops/--ks-bytes (plus --ks-calibration machine-profile.json for measured peaks)
    sm__throughput.avg.pct_of_peak_sustained_elapsed            n/a          [unavailable]  requires GPU hardware performance counters, which are restricted here (RmProfilingAdminOnly=1 and no CAP_SYS_ADMIN/CAP_PERFMON in the container); CUPTI reports 

----------------------------------------------------------------------------------------------------
Kernel: (device)
  0 launch(es), 0.0000 ms total, 0.0000 ms mean

  Section: DeviceStatistics                   [estimated]
    meas__nvml_mem_busy_mean_pct                               25 %        [measured]  [resolution 1 %（整数百分比）]  (区间平均；是「显存子系统忙碌占比」，不是 DRAM 吞吐本身。读数 0.0% 只说明低于分辨率下限，不等于「DRAM 完全没动」；必须让 GPU 保持饱和，否则会被宿主间隙稀释（实测 96.8% vs 6.3%）)
    est__dram_busy_equivalent_gbs                           227.3 GB/s     [estimated]  [resolution 1 %]  (等效带宽口径；与按字节/时间的口径略有差异)
    meas__sm_clock_mean_mhz                                1762.5 MHz      [measured]  [resolution 1 MHz / 1 W（驱动报数粒度）]
    meas__sm_clock_max_mhz                                   2790 MHz      [measured]  [resolution 1 MHz / 1 W（驱动报数粒度）]
    meas__power_max_w                                       220.9 W        [measured]  [resolution 1 MHz / 1 W（驱动报数粒度）]
    meas__device_energy_j                                 414.921 J        [measured]  [resolution 1 mJ]  (ncu 不提供能耗指标)

  Section: DramTraffic                        [estimated] (+2 n/a)
    est__dram_bytes_total                                  909.78 GB       [estimated]  [resolution 1 %]  (窗口已按首个/末个忙碌采样点裁到稳态段（--ks-window auto）：参与积分 1.4 s，裁掉约 2.8 s 启动/JIT/空闲；用 --ks-window full 可关闭；窗口覆盖率 71%（SM≥5% 或显存≥20% 的采样点占比）；采样覆盖时间不足窗口一半：采样器可能启动过晚或提前退出；采样周期 200 ms、区间平均（utilization.memory 分辨率 1 个百分点）；字节 = Σ(mem_util×Δt)×实测峰值带宽)
    est__dram_bytes_per_launch                                n/a          [unavailable]  窗口覆盖率 71% < 80%：无法把 909783 MB 总量归因到每个 launch；把稳态循环延长到约 6 秒，或加 --ks-window auto
    est__dram_throughput_pct_of_peak                         71.4 %        [estimated]  [resolution 1 %]  (忙碌时间占比口径，与按字节/时间的口径略有差异)
    ks__sampler_window_coverage_pct                          71.4 %        [derived]  [resolution 1 个采样点]  (前提①「窗口被被测负载主导」：<80% 不给 per-launch 归因，<50% 连总量也不给)
    est__dram_bytes_per_launch_trace                          n/a          [unavailable]  未提供访存轨迹：加 --ks-memtrace <trace> 可换用轨迹口径，在 NVML 代理口径不成立时仍能完成账本对拍

----------------------------------------------------------------------------------------------------

Provenance summary: exact=18, derived=1, measured=5, estimated=3, unavailable=5
Legend: exact=与 ncu 同源 | derived=本机实测标定/模型 | measured=我们的探针实测 | estimated=采样或代理(附误差) | unavailable=需要硬件计数器，只报 n/a
Output(json): bench/artifacts/20261008-225555/rep-compute.ksreport.json
Output(perfetto): bench/artifacts/20261008-225555/rep-compute.perfetto.json
Open the timeline at https://ui.perfetto.dev -> Load trace -> the file above
